Encrypted login | home

Program Information

Extension of the Fermi-Eyges Most Likely Path in Heterogeneous Medium with Prior Knowledge Information

no image available
C Collins-Fekete

C-A Collins-Fekete1,2,3*, E Baer4 , L Volz3 , H Bouchard5 , L Beaulieu1.2 , J Seco3 , (1) Universite Laval, Quebec, QC, (2) Centre hospitalier universitaire de Quebec, Quebec, QC, (3) DKFZ Heidelberg, Heidelberg,Baden-Wurttemberg, (4) UCL, London, UK, (5) Universite de Montreal, Montreal, Quebec,


SU-H2-GePD-J(A)-4 (Sunday, July 30, 2017) 3:30 PM - 4:00 PM Room: Joint Imaging-Therapy ePoster Lounge - A

Purpose: Particle imaging suffers from poor spatial resolution due to the multiple Coulomb scattering deflections undergone by the particles throughout their path. A most-likely path (MLP) formalism was developed to account for these deflections based on the Fermi-Eyges theory combined with the Highland scattering power. Initial applications calculate the MLP formalism in a homogeneous water medium. However, this potentially reduces the accuracy of the MLP estimate as well as the achievable resolution of the subsequent tomographic reconstruction. This work intends to investigate a possible gain in precision by introducing prior knowledge on the medium radiation length and stopping power in the MLP formalism.

Methods: The Monte Carlo (MC) Geant4 algorithm is used to simulate protons (n=10⁶) crossing two parametric phantoms representing the lung and abdomen regions. The prior knowledge information is gathered from 1) the MC simulation for prior-knowledge ground-truth (MLP-Geant4) and from 2) a recent DECT material decomposition technique for a realistic application (MLP-DECT). The reconstructed path accuracy using prior-knowledge methods is compared to 3) the homogeneous water scenario (MLP-Water) and 4) a scenario where the proton path is projected linearly up to a Hull at the boundary of the phantom (MLP-Hull) to account for air gaps. The maximal root-mean-square error (RMSmax) between the reconstructed path and the MC path is compared for each method.

Results: In the lung, the RMSmax is decreased between the MLP-Water (0.54mm) and the three other scenarios (0.34mm), with no significant differences between them. In the abdomen, the RMSmax is decreased between the MLP-Water scenario (0.56 mm) and the MLP-Hull (0.30mm), but not significantly further with the MLP-DECT/MLP-Geant4 (0.29mm) scenarios.

Conclusion: The introduction of prior-knowledge in the MLP formalism decreases the RMS error to the MC path, but no further than a simpler Hull algorithm. The Hull is suggested for future particle imaging applications.

Contact Email: